Am. J. Trop. Med. Hyg., 96(6), 2017, pp.

1404–1414
doi:10.4269/ajtmh.16-0170
Copyright © 2017 by The American Society of Tropical Medicine and Hygiene

The Influence of Household- and Community-Level Sanitation and Fecal Sludge Management
on Urban Fecal Contamination in Households and Drains and Enteric Infection in Children
David Berendes,1,2* Amy Kirby,2,3 Julie A. Clennon,2,4 Suraja Raj,2 Habib Yakubu,2 Juan Leon,2,3 Katharine Robb,2
Arun Kartikeyan,5 Priya Hemavathy,5 Annai Gunasekaran,5 Ben Ghale,5 J. Senthil Kumar,6
Venkata Raghava Mohan,6 Gagandeep Kang,5 and Christine Moe2,3
1
Department of Environmental Engineering, School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia;
2
Center for Global Safe Water, Sanitation, and Hygiene, Rollins School of Public Health, Atlanta, Georgia; 3Hubert Department of Global
Health, Rollins School of Public Health, Emory University, Atlanta, Georgia; 4Department of Biostatistics and Bioinformatics, Rollins School
of Public Health, Emory University, Atlanta, Georgia; 5Wellcome Research Laboratory, Christian Medical College, Vellore, India;
6
Department of Community Health, Christian Medical College, Vellore, India

Abstract. Urban sanitation necessitates management of fecal sludge inside and outside the household. This study
examined associations between household sanitation, fecal contamination, and enteric infection in two low-income
neighborhoods in Vellore, India. Surveys and spatial analysis assessed the presence and clustering of toilets and fecal
sludge management (FSM) practices in 200 households. Fecal contamination was measured in environmental samples
from 50 households and household drains. Enteric infection was assessed from stool specimens from children under
5 years of age in these households. The two neighborhoods differed significantly in toilet coverage (78% versus 33%)
and spatial clustering. Overall, 49% of toilets discharged directly into open drains (“poor FSM”). Children in households
with poor FSM had 3.78 times higher prevalence of enteric infection when compared with children in other households,
even those without toilets. In the neighborhood with high coverage of household toilets, children in households with
poor FSM had 10 times higher prevalence of enteric infection than other children in the neighborhood and drains in
poor FSM clusters who had significantly higher concentrations of genogroup II norovirus. Conversely, children in
households with a toilet that contained excreta in a tank onsite had 55% lower prevalence of enteric infection com-
pared with the rest of the study area. Notably, households with a toilet in the neighborhood with low toilet coverage
had more fecal contamination on floors where children played compared with those without a toilet. Overall, both toilet
coverage levels and FSM were associated with environmental fecal contamination and, subsequently, enteric infection
prevalence in this urban setting.

INTRODUCTION tamination, with concurrent changes in sanitation cover-

age.12 While several quantitative microbial risk assessments
Poor water, sanitation, and hygiene are associated with (QMRAs) have modeled environmental transmission path-
multiple adverse health and developmental outcomes.1,2 ways for exposure to fecal contamination and risk of enteric
Following the Millennium Development Goals, the focus of infection, few studies have measured associations between
the sanitation sector has been in rural areas, where 70% of microbial indicators or pathogens in the household or public
those without access to improved sanitation live.3 However, environments and type or coverage of sanitation.14,15 Because
the need for sanitation solutions in poor urban neighbor- effective sanitation is expected to decrease enteric infection
hoods and informal settlements is a growing concern, as risk through safe containment of excreta, it is important to
the world’s population has shifted to being predominantly examine changes in environmental fecal contamination as an
urban within the last decade.4,5 By 2050, the urban popula- intermediate outcome.16
tion is estimated to almost double, from 3.3 to 6.3 billion, Systems-level approaches to urban sanitation, where
providing a new challenge for sanitation implementers and containment of excreta requires more than a household toilet
policymakers.4 Within poor, dense urban neighborhoods alone, have not been well examined.5 When compared with
with little sanitation infrastructure, frequent person-to-person rural settings, urban sanitation presents complex challenges,
contact and poor environmental conditions facilitate trans- in particular the spillover of fecal contamination from private
mission of fecal–oral infections, yielding frequent diarrhea in to public domains and vice versa. Preventing this spillover
young children.4,6–11 requires consideration of the entire sanitation chain to ensure
Despite links between poor sanitation and health, evidence safe containment, transport, treatment, and ultimately dis-
of effective sanitation in urban settings remains weak. posal or reuse of excreta.17 Components of the sanitation
Recent meta-analyses have identified few interventions in chain start with the user interface (household toilet), but also
urban neighborhoods and limited evidence, overall, of the include transport (e.g., sewerage or onsite containment
positive effects of sanitation on diarrheal disease.12,13 Among followed by emptying and trucking of fecal sludge) and even-
other limitations, the authors highlight bias in outcome mea- tual treatment of the fecal sludge. These components beyond
sures (self-reported diarrhea) and poor mechanistic evidence the toilet are encompassed by the current focus on “fecal
of changes in more proximal exposures, such as fecal con- sludge management” (FSM).18 To date, associations between
urban FSM, fecal contamination, and adverse health out-
comes have only been evaluated for sewerage interventions,
*Address correspondence to David Berendes, Department of which were associated with reduced diarrhea incidence.19
Environmental Engineering, School of Civil and Environmental
Engineering, Georgia Institute of Technology, Ford ES&T 3368, 311 Because sewerage may not be feasible in some urban set-
Ferst Dr., Atlanta, GA 30332-0002. E-mail: david.berendes@ce. tings and because many current sewer and open drain con-
gatech.edu nections do not result in treatment of the fecal sludge, it is
1404
 URBAN SANITATION AND FECAL CONTAMINATION 1405

important to examine the effects of other FSM models, hold (household toilets), public toilets, and open defecation.
including onsite containment, on health outcomes.6 Water sources were predominantly municipal taps, providing
In addition to the linear sanitation chain, studies must intermittent water supply for a few days each week, thus
also consider the spatial heterogeneity of urban sanitation water was stored at the home as well. Water treatment (boil-
coverage. While the effect of toilets on environmental fecal ing) was practiced by some of the population and was
contamination and enteric infection is often measured at the recorded as part of the CMC hygiene survey (below and
household level, sanitation may have community-level bene- Supplemental Information).
fits in areas of high sanitation coverage, even for residents In each neighborhood, environmental samples and stool
without a toilet.20–22 Because of the interconnectedness of specimens were collected from 25 households selected
public and private urban environments, there is a growing 1) from neighborhood sampling frames from previous CMC
need to examine how the concentrations of fecal contami- studies and 2) based on the score from a hygiene survey
nation in the environment vary with the underlying spatial developed by CMC, previously validated, and implemented
distribution and clustering of sanitation in the community.23 1 month before SaniPath data collection.28,34,34 Briefly, this
While management of fecal sludge in low-income, urban survey assessed 18 household hygiene characteristics and
areas is generally poor, the associations between FSM and behaviors related to water collection, infant cleaning and
fecal contamination or enteric infection within that environ- feeding practices, and defecation, using Yes (equal to 1) or
ment have not been examined.7 There is a need to under- No (equal to 0) responses to create an additive household
stand how fecal contamination and enteric infection vary with hygiene score. Scores less than or equal to 9 were classified
toilets and FSM in the urban environment. This study exam- as “poor” hygiene, whereas scores greater than 9 were clas-
ines the associations between household sanitation, includ- sified as “good” hygiene. The hygiene survey was adminis-
ing toilets, their associated FSM, and their spatial clustering, tered to all households in each study neighborhood before
and fecal contamination within the household and nearby the study’s commencement. To ensure variation in general
drains, as well as pediatric enteric infection. By assessing household-level hygiene practices, 13–14 poor hygiene
proximal exposure outcomes and more distal infection out- (those with the lowest scores in the survey) and 11–12 good
comes in two different urban neighborhood environments, hygiene households (those with the highest scores in the
this work will contribute to understanding how the urban survey) were selected randomly in each neighborhood. Fur-
environment affects the success of sanitation interventions. ther detail, including the survey instrument, can be found in
the Supplemental Information.
MATERIALS AND METHODS Ethical approval. Approval was obtained from the Emory
University Institutional Review Board (IRB) and the CMC IRB.
Data source. This study was conducted as a sub-study Informed consent was obtained before sample collection
of a SaniPath Exposure Assessment Tool24 deployment in and survey administration at each household.
two low-income urban neighborhoods in Vellore, India. Data Environmental and stool sample collection, analysis,
were collected in February–March and September 2014. and processing. Environmental samples—including hand
Study site. The study site consisted of two neighborhoods rinses from children less than 5 years of age, rinses of a
in Vellore, Chinnallapuram and Old Town, which were chosen sentinel object, swabs of household floors, and 500 mL of
because of their low socioeconomic status, poor sanitation, drain water—and stool specimens from children under 5 years
and long-standing relationship with the Christian Medical of age were collected in March 2014. After collection, sam-
College (CMC), which provided access to spatial and demo- ples and specimens were stored on ice for up to 4 hours until
graphic data from previous studies.25,26 The Old Town arrival at the laboratory, where they were refrigerated at 4°C.
neighborhood is the site for the Interactions of Malnutrition Environmental samples were analyzed for Escherichia coli
and Enteric Infections: Consequences for Child Health and within 6 hours of receipt by membrane filtration and plating
Development (MAL-ED) study in Vellore.25 Of the seven con- on m-ColiBlue24® Medium (Hach Company, Loveland, CO)
tiguous sub-neighborhood areas selected for the MAL-ED according to the U.S. Environmental Protection Agency method
study, five were selected for the SaniPath Tool deployment. 1604.36 Enteroaggregative E. coli (EAEC) and genogroup I
The Chinnallapuram study area is a 0.41 km2 semi-urban neigh- and II (GI and GII) norovirus were assessed by quantitative
borhood with a reported population density of 30,520/km2.27 polymerase chain reaction (PCR) and reverse transcription
Old Town is a 0.33 km2 urban neighborhood with an esti- PCR (RT-PCR). Escherichia coli was chosen as an indicator
mated population density of 41,977/km2.25 The study area of fecal contamination, whereas EAEC and norovirus were
within Old Town was approximately 0.18 km2. Vellore is sub- chosen based on their high prevalence in children of the
ject to two monsoon seasons (a southwest monsoon from Vellore field site for the MAL-ED study.37,38 Further, both are
June to September and a northeast monsoon from October predominantly human-specific infections.39
to December), with the remaining January to May period as a Stool specimens were analyzed for enteropathogens
dry season.25 using the MAL-ED study protocols, with the exception of
Among the study population, houses had floors made of Campylobacter spp., which was assessed by PCR.29,30,38
cement or concrete (84%), ceramic tiles (10%), or earth or Further information on pathogens analyzed is provided in
mud (6%). Walls were mostly cement or concrete also (81%), the Supplemental Information.
though 15% had mud walls.25 Households were mostly Hand rinses. At each household, a hand rinse sample was
Hindu (59.9%) or Muslim (33.7%) while few were Christians collected from the child under 5 years of age who was previ-
(6.4%, data from a local census maintained by the Depart- ously enrolled in other CMC studies. The child’s right hand
ment of Gastrointestinal Sciences at CMC). Sanitation was inserted into a sterile 2-L Whirl-Pak bag (Nasco, Fort
consisted of pour-flush toilets with slabs within the house- Atkinson, WI) containing 500 mL of sterile phosphate-buffered
1406 BERENDES AND OTHERS

saline (PBS) solution. The staff massaged the fingers and ment of potential PCR inhibition. Any samples that were
palm for 30 seconds in the PBS, then the child removed positive (at least one well with a cycle threshold (Ct) value
their right hand, inserted their left hand, and the massage less than 45) or inhibited were quantified using the OneStep
procedure was repeated. At the laboratory, the sample was PCR (EAEC) or RT-PCR (norovirus) kit (Qiagen) and a
diluted 1:100, 1:101, and 1:102 in sterile PBS before mem- standard curve. Positive and negative controls for EAEC or
brane filtration. Before PCR, 200 mL of the original hand norovirus were included with every PCR run.
rinse sample was precipitated with 12% polyethylene glycol Samples tested for GI or GII norovirus using the OneStep
(PEG) 8000, centrifuged for 20 minutes at 6,000 rpm, and Kit and classified as positive (both wells had Ct values less
suspended in 5 mL sterile water, of which 1.5 mL was further than or equal to 45 and a difference of less than or equal to
concentrated by precipitation with 12% PEG 8000 before 4 between Ct values for duplicate wells) were quantified by a
nucleic acid extraction.40 simple average of both wells. Due to inconsistent standard
Sentinel objects. At each household, a child’s toy or feed- curves affecting quantification—but not detection—of EAEC,
ing spoon, volunteered by the mother, was used as the samples tested for EAEC using the OneStep kit were classi-
“sentinel object.” The object was inserted into a sterile 2-L fied as positive or negative. Samples with no detectable
Whirl-Pak bag containing 500 mL of sterile PBS, massaged EAEC or norovirus were assigned the value of the theoretical
from the outside of the bag for 1 minute, and subsequently lower limit of detection for the assay (334 cell equivalents [CE]
removed and returned to the family. Rinses from sentinel for EAEC or genome equivalent copies (GEC) for norovirus
objects were processed identically to hand rinse samples GI and GII per 100 mL (2.52 log10 CE or GEC/100 mL)).
for membrane filtration and PCR. Survey data collection. In each study neighborhood,
Household floor swabs. At each household, a composite surveys were conducted in 100 households: 25 households
household floor swab was collected using EnviroMax Plus were those with concurrent environmental sample and stool
Sterile Environmental Swabs (Puritan Medical Products, specimen collection, whereas the remaining 75 were divided
Guilford, ME). The child’s play area was identified by the equally across sub-neighborhood areas and chosen at ran-
mother, after which the field staff used a 24 × 16-cm framing dom within them. To be eligible, households had to have
square to outline a 25 cm2 area in each of four corners and a child under 5 years of age. The target respondent for
the center of the play area and swabbed back and forth the survey was the person responsible for water, sanitation,
hygiene, and food activities, generally the mother of the
across those sections of the floor. Two swabs were used to
youngest child, or rarely, the grandmother. If the respondent
cover the entire area and subsequently combined into a
was not available and the household was one of the house-
single sample covering a total surface area of 125 cm2. Each
holds where environmental stool samples were to be col-
swab was eluted in 7 mL of PBS solution in a sterile con-
lected, survey enumerators returned to the household at a
tainer, and the eluates from both swabs were combined for
later time, otherwise, the nearest available household was
an approximate sample volume of 14 mL. From this volume,
selected for survey. The household survey included ques-
dilutions of 1:100, 1:101, and 1:102 were made and mem-
tions about the household’s population, presence of a toilet,
brane filtered. Nucleic acids were extracted from 1.5 mL of
and FSM practices, as well as the children’s and adult’s def-
swab eluate after one round of PEG precipitation.
ecation practices. A Global Positioning System location was
Drain water. A sample of drain water was collected from collected at each household using Garmin eTrex Venture HC
the drain directly in front of the household, regardless of devices (Garmin International Incorporated, Olathe, KS).
the discharge location of the household toilet (if applicable). Analyses. Concentrations per pair of hands, sentinel object,
A sterile bailer or stainless steel ladle was used to collect and 125 cm2 of household floor were back-calculated using
approximately 500 mL of drain water into a sterile 2-L the rinse volume for these samples. All microbial concentra-
Whirl-Pak bag, taking care not to disturb sediment on the tions were log10-transformed before statistical analyses.
bottom or nearby trash. Drain samples were diluted 1:101, Values below the lower detection limit for membrane filtration
1:102, and 1:103 in sterile PBS at the laboratory before were substituted on the log10 scale with the value of the lower
membrane filtration. DNA and RNA were extracted from limit of detection (1 colony-forming unit [CFU]/100 mL),
1.5 mL samples of the original sample before PCR analysis. accounting for sample dilution.
Because almost all drain water samples collected during Statistical analyses were conducted in R version 3.2.3
the initial (February–March) sampling period had E. coli col- (R Foundation for Statistical Computing, Vienna, Austria) using
ony numbers above the countable range on the filter mem- base packages and the “lme4” package for mixed-effects
brane, 10 of the original 25 households in each neighborhood logistic regression.41–44 Linear regression was used to assess
were resampled spatially at random in September 2014, per continuous outcomes (E. coli concentrations in all environmen-
the original sampling protocol. These samples were analyzed tal samples and norovirus GII concentrations in drain samples)
by membrane filtration for E. coli after 1:104–1:106 dilutions. while logistic regression was used to assess binary outcomes
Stool specimens. A stool specimen was collected from (presence/absence of EAEC and norovirus GII in drain sam-
each child under 5 years of age in the 50 study households. ples, as well as pathogen detection in stool specimens).
Further detail on the analysis methods is given in the Sup- Binary (presence/absence) data from household surveys
plemental Information. were evaluated for most-likely local clustering in SaTScan
Quantitative real-time PCR. Further detail on probes, version 9.4 (SaTScan.com, Boston, MA) using Kulldorff’s
primers, and standard curves used in quantitative PCR Bernoulli spatial scan, which evaluates binary outcomes in
analyses can be found in the Supplemental Information. All point data distributed in space to assess the degree of
samples were tested using the Qiagen QuantiFast Pathogen + nonrandom clustering of “0” or “1” values.45 An α of 0.05 was
IC Kit (Qiagen Sciences, Germantown, MD) (PCR for EAEC used to determine significance in cluster analysis and regres-
and RT-PCR for norovirus) for initial screening and assess- sion modeling.
 URBAN SANITATION AND FECAL CONTAMINATION 1407

RESULTS poor FSM was 78% in the significant cluster of poor FSM
present. In Old Town, both significant clusters were small,
Frequency and within-neighborhood spatial clustering but had 100% coverage of households with toilets with poor
of household sanitation. To compare sanitation coverage FSM. No significant clusters of high coverage of toilets with
and spatial heterogeneity within and between study neigh- good FSM were detected.
borhoods, we assessed the frequency and type of house- Microbiological concentrations in environmental
hold toilets and FSM and their most-likely clustering in samples. Levels of fecal contamination in the household
Chinnallapuram and Old Town (Table 1, Figures 1 and 2). In were characterized by examining rinses of children’s hands,
both neighborhoods, toilets either discharged directly to an rinses of sentinel objects, and swabs of household floors.
open drain (defined as “poor FSM”) or were connected to a tank Escherichia coli, EAEC, and GI and GII norovirus levels were
under the household that contained the excreta (defined as quantified in these samples. Distributions of sample E. coli
“good FSM”). Compared with households in Chinnallapuram, concentrations were all approximately normal when log-
those in Old Town reported a significantly lower proportion transformed, though sentinel object rinses did have frequent
of household toilets (33% versus 78%), and more household left-censored values from non-detects (data not shown).
toilets had poor FSM (82% versus 35%). Among households Escherichia coli were detected in most hand rinse, sentinel
with toilets, reported use of toilets was high in Chinnallapuram object, and household swab samples (Table 2: Overall). To
(76/78 households) and Old Town (32/33 households, data not assess differences in fecal contamination within households
shown). Open defecation was more common in Old Town than by neighborhood and household hygiene practices, logistic
in Chinnallapuram, but use of public toilets was similar (54% and linear regression models were constructed for E. coli
and 59%, P = 0.57). Frequent use of public toilets (more than data (Table 2: Associations with neighborhood and hygiene
10 times per month) was more common in Old Town (18%) status). Overall, there was no significant variation in E. coli
than Chinnallapuram (4%, P < 0.01). detection or concentrations in samples by neighborhood or
Significant clustering of high coverage of household toilets hygiene status.
was present in Chinnallapuram, but not Old Town. Signifi- EAEC was detected in 1/50 hand rinses, 0/50 sentinel
cant spatial clusters of low coverage of household toilets objects rinses, and 1/50 floor swabs. GI norovirus was not
and high coverage of household toilets with poor FSM were detected in samples within the household. GII norovirus was
present in both neighborhoods (Table 1: Most likely cluster, detected in 1/50 hand rinse samples, 0/50 sentinel object
Figures 1 and 2). Clusters of low coverage of household toi- rinses, and 0/50 floor swabs. These EAEC and GI and GII
lets varied in size (40 households in Chinnallapuram versus norovirus samples were omitted from further analysis due to
27 households in Old Town) and coverage within the cluster low levels of detection.
(50% in Chinnallapuram versus 0% in Old Town). In Levels of fecal contamination outside the household
Chinnallapuram, coverage of households with toilets with were characterized by examining drain samples, which were

TABLE 1
Reported frequency and clustering of household sanitation and FSM in Chinnallapuram and Old Town
Chinnallapuram (N = 100) Old Town (N = 100) Overall (N = 200)
Count (%) Count (%) Count (%) P value*

Household-level
Household toilet† 78 (78.0) 33 (33.0)‡ 111 (55.5) < 0.01
FSM: Toilet excreta contained onsite§ 37 (47.4) 3 (9.1) 40 (36.0) < 0.01
FSM: Toilet discharges directly to drain§ 27 (34.6) 27 (81.8) 54 (48.6) < 0.01
FSM: Other/do not know§ 14 (18.0) 2 (6.1) 16 (14.4) 0.18
Open defecation
< 5-year-olds 40 (40.0) 80 (80.0) 120 (60.0) < 0.01
Respondent (adult) 19 (19.0) 68 (68.0) 87 (43.5) < 0.01
Public toilet use (by respondent)
None 41 (41.0) 46 (46.0) 87 (43.5) 0.57
Low (1–5 times per month) 51 (51.0) 31 (31.0) 82 (41.0) 0.01
Medium (6–10 times per month) 4 (4.0) 5 (5.0) 9 (4.5) > 0.99
High (> 10 times per month) 4 (4.0) 18 (18.0) 22 (11.0) < 0.01

Chinnallapuram Old Town

Count (cluster prevalence) P value∥ Count (cluster prevalence) P value∥

Most likely clusters¶

Household toilet
High-coverage cluster 43 (100.0) < 0.01 – –
Low-coverage cluster 40 (50.0) < 0.01 27 (0.0) 0.02
FSM: toilet discharges directly to drain
High-coverage cluster 18 (77.8) 0.01 9 (100.0), 7 (100.0) 0.02, 0.04
FSM = fecal sludge management.
*P value for t test of proportions between neighborhoods.
†All toilets were pour-flush toilets.
‡Of the 33 households reporting having a toilet, 32 responded to the subsequent questions about FSM.
§Percent in parentheses represents the percentage of all households with toilets.
¶No significant clusters of households with toilet excreta contained onsite (good FSM) were observed.
∥P value for comparison of the prevalence of the attribute within the cluster compared with the overall prevalence of the attribute in the neighborhood. Only clusters significant at the 0.05
level are presented, otherwise “–” is presented.
1408 BERENDES AND OTHERS

FIGURE 1. Sanitation coverage and clustering in Chinnallapuram.

analyzed for E. coli, EAEC, and GI and GII noroviruses. (1/76), Giardia spp. (17/76), GII norovirus (5/76), and patho-
Escherichia coli was detected in 50/50 drain samples collected genic E. coli (14/76) were also detected (data not shown).
initially (February–March 2014), with concentrations above Variation in enteric pathogen prevalence by neighborhood
the detection limit in 49/50 samples and in 20/20 resamples and household hygiene status was assessed using mixed-
(September 2014), with concentrations above the detection effects logistic regression for pathogens detected frequency
limit in 1/20 samples, a mean of 6.83 log10 CFU/100 mL (> 20% of stool specimens, Table 3). The prevalence of any
(standard deviation [SD]: 0.56 log10 CFU/100 mL), and an enteric pathogens in stool, as well as that of specific patho-
approximately normal distribution when log-transformed. gens, did not vary significantly by neighborhood or house-
Because drain resampling took place during a different season hold hygiene status, with the exception of Campylobacter
than the rest of data collection, drain E. coli concentrations spp. detection. Households with poor hygiene had signifi-
were excluded from further analyses. EAEC, GI norovirus, cantly higher detection of Campylobacter spp. in stool com-
and GII norovirus were detected in 15/50, 1/50, and 19/50 pared with those with good hygiene (prevalence ratio: 3.42,
drain samples, respectively. Mean concentrations of EAEC P = 0.02). Because no single pathogen was associated with
and GII norovirus were 2.67 log10 CE/100 mL (SD: 0.41 log10 more than half of infections, further analyses were limited to
CE/100 mL) and 3.43 log10 GEC/100 mL (SD: 1.41 log10 the presence of any enteric infection (i.e., pooled pathogens)
GEC/100 mL), respectively. as an outcome.
Enteric pathogen detection in stool. To determine the Association between household- and cluster-level
prevalence of enteric infections in children, a stool specimen toilet coverage, FSM practices, and within-household
from each child under 5 years of age in the 50 study house- fecal contamination. Associations between household toilet
holds where environmental samples were collected was presence, FSM practices, and within-household fecal con-
assayed for viral, bacterial, protozoan, and parasitic enteric tamination were examined using multivariate linear regression
pathogens. Overall, one or more enteric pathogens was (Supplemental Tables 1–3). No significant associations were
detected in 51/76 stool specimens (67%). Campylobacter observed between household- or cluster-level sanitation
spp. was most frequently detected in stool specimens (32/ variables and E. coli concentrations on children’s hands (Sup-
76 children), but astrovirus (7/76), Entamoeba histolytica plemental Table 1).
 URBAN SANITATION AND FECAL CONTAMINATION 1409

FIGURE 2. Sanitation coverage and clustering in Old Town.

Associations between E. coli concentrations on floors and E. coli concentrations in floor swabs in households with
presence of a household toilet varied by neighborhood and without a toilet in Chinnallapuram. However, in Old
(Supplemental Table 2). At the household-level (ignoring Town, the presence of a toilet was associated with higher
spatial clustering), there was no significant difference in E. coli concentrations on floors (difference of 1.15 log10

TABLE 2
Variation in detection and concentrations of Escherichia coli in environmental samples within households with neighborhood and hygiene status
Child hand rinse (N = 50) Sentinel object rinses (N = 49)* Household swabs (N = 50)†

E. coli detection E. coli concentration E. coli detection E. coli concentration E. coli detection E. coli concentration

Number of samples Geometric mean (SD)‡ Number of samples Geometric mean (SD)§ Number of samples Geometric mean (SD)¶

Overall 45/50 107.2 (11.7) 32/49 13.2 (5.9) 48/50 245.5 (9.8)
Associations with neighborhood and hygiene status
OR (95% CI)∥ β** SE(β) P value OR (95% CI)∥ ↆ SE(β) P value OR (95% CI)∥ ⇇ SE(β) P value

Neighborhood§§ 0.22 (0.01, 1.62) 0.51 0.30 0.09 1.27 (0.39, 4.22) 0.15 0.22 0.51 1.00 (0.04, 26.3) 0.07 0.28 0.81
Poor hygiene¶¶ 0.18 (0.01, 1.36) −0.07 0.31 0.83 0.99 (0.30, 3.28) 0.19 0.22 0.40 0.85 (0.03, 22.2) −0.28 0.28 0.33
Neighborhood§§ 0.19 (0.01, 1.48) 0.51 0.30 0.10 1.28 (0.39, 4.23) 0.16 0.22 0.48 0.99 (0.04, 26.1) 0.06 0.28 0.84
Poor hygiene¶¶ 0.16 (0.01, 1.24) −0.05 0.30 0.88 1.01 (0.31, 3.34) 0.20 0.22 0.38 0.85 (0.03, 22.3) −0.28 0.28 0.34
CI = confidence interval; OR = odds ratio; SD = standard deviation; SE = standard error.
*One sentinel object rinse was unable to be read and thus was not included in the results. Sentinel objects were plastic (28/50), metal (17/50), other material (4/50), or mixed material (1/50).
†Household floors were cement (39/50), tile (8/50), or other material (3/50).
‡Units are colony-forming unit (CFU)/pair of hands.
§Units are CFU/100 mL.
¶Units are CFU/125 cm2.
∥Though P values for ORs are omitted for reasons of space in the table, none were significant at α = 0.05.
**Estimate is in log10 CFU/pair of hands.
††Estimate is in log10 CFU/100 mL.
‡‡Estimate is in log10 CFU/125 cm2.
§§Old Town neighborhood (reference is Chinnallapuram).
¶¶Hygiene status was divided into “poor” or “good” hygiene categories based on a 18-point scale (0–9 as “poor,” 10–18 as “good”) discussed in Methods section and presented in Collinet-
Adler and others.34
1410 BERENDES AND OTHERS

TABLE 3
Variation in detection of enteric pathogens in stool with neighborhood and hygiene status*
Any enteric pathogen Campylobacter spp. Giardia spp. Pathogenic Escherichia coli†

PR (95% CI) P value PR (95% CI) P value PR (95% CI) P value PR (95% CI) P value

Neighborhood: Old Town 1.32 (0.50, 3.49) 0.57 1.91 (0.71, 5.14) 0.18 1.56 (0.48, 6.46) 0.45 0.73 (0.14, 3.78) 0.64
Poor hygiene 1.97 (0.75, 5.62) 0.17 3.42 (1.30, 12.3) 0.02 1.69, (0.53, 7.27) 0.39 0.55 (0.07, 3.37) 0.34
Neighborhood: Old Town 1.37 (0.51, 3.72) 0.53 2.15 (0.80, 5.97) 0.13 1.60 (0.49, 6.78) 0.47 0.70 (0.11, 5.02) 0.56
Poor hygiene 2.01 (0.76, 5.77) 0.16 3.61 (1.37, 11.4)† 0.01 1.72 (0.54, 7.57) 0.54 0.53 (0.09, 4.10) 0.31
CI = confidence interval; PR = prevalence ratio for detection of enteric pathogen in stool specimen.
*N = 76 children from which stool specimens were collected (43 in Chinnallapuram, 33 in Old Town). Enteric pathogens detected in stool specimens included astrovirus, Campylobacter
spp., Entamoeba histolytica, Giardia spp., genotype II norovirus, and pathogenic E. coli. A full list of organisms tested in stool specimens is presented in Houpt and others.35 Only pathogens
detected in > 20% of stool specimens were regressed against neighborhood and hygiene status.
†Enteroaggregative E. coli, enterohemorrhagic E. coli, enteropathogenic E. coli, and enterotoxigenic E. coli.

CFU/125 cm2, P = 0.06), compared with households with- models for EAEC and both logistic and linear regression
out a toilet (Supplemental Table 2: Household level). At the models for GII norovirus in drains were constructed.
cluster level, households in the cluster of low coverage of Though significant associations were not detected for
household toilets in Old Town (0% coverage within the EAEC (data not shown), large effect sizes were observed
cluster) had significantly lower E. coli concentrations on that may be worthy of future investigation. For example,
floors compared with the rest of the neighborhood (differ- EAEC was less likely to be detected in drains outside
ence of −1.38 log10 CFU/125 cm2, P = 0.02, Supplemental households with poor, compared with those with good,
Table 2: Most likely clusters). No difference in E. coli con- hygiene scores (odds ratio [OR]: 0.30, 95% confidence
centrations was observed within and outside the cluster of interval [CI]: 0.08–1.02, P = 0.06). Further, drain samples in
low coverage of household toilets in Chinnallapuram (50% clusters of low coverage of household toilets were less
coverage within the cluster). likely to be positive for EAEC (OR: 0.27, 95% CI: 0.05–1.14,
Escherichia coli concentrations in rinses of sentinel objects P = 0.09), whereas those in clusters of high coverage of
were not significantly associated with any sanitation vari- household toilets were more likely to be positive (OR: 5.01,
ables (Supplemental Table 3). 95% CI: 0.70–47.6, P = 0.12).
Associations between demographics, neighborhood, GII norovirus detection and concentrations in drains did
and household- and cluster-level sanitation variables not vary significantly by neighborhood or hygiene status
and fecal contamination outside the household. To exam- (Table 4: Demographics and hygiene). At the household-
ine variation in pathogen levels in the public domain with level, the odds of detection and concentrations of GII
household and neighborhood sanitation, logistic regression norovirus in drain samples were higher for drains adjacent

TABLE 4
Variation in GII norovirus detection and concentration in drain water
GII norovirus detection GII norovirus concentration

OR (95% CI)* P value β† SE(β) P value

Demographics and hygiene

Neighborhood: Old Town 0.42 (0.13, 1.34) 0.15 −0.38 0.39 0.35
Poor hygiene 0.92 (0.29, 2.91) 0.88 −0.05 0.40 0.90
Neighborhood: Old Town 0.42 (0.12, 1.33) 0.15 −0.38 0.40 0.35
Poor hygiene 0.88 (0.27, 2.86) 0.83 −0.07 0.41 0.87
Sanitation (household level)
Household toilet 4.73 (0.93, 28.9) 0.07 0.90 0.53 0.10
Toilet excreta contained onsite‡ 1.51 (0.37, 6.10) 0.56 0.41 0.49 0.41
Toilet discharges to drain‡ 1.14 (0.28, 4.74) 0.85 0.17 0.50 0.74
Toilet excreta contained onsite§ 1.81 (0.38, 8.95) 0.45 0.61 0.55 0.28
Toilet discharges to drain§ 1.50 (0.31, 7.62) 0.61 0.44 0.55 0.43
Open defecation (< 5-year-old) 0.07 (0.01, 0.58) 0.02 −0.95 0.52 0.07
Open defecation (< 5-year-old)/OT 0.69 (0.05, 9.67)¶ 0.09¶ NI∥ – –
Open defecation (adult) 0.08 (< 0.01, 0.78) 0.05 −0.62 0.58 0.30
Open defecation (adult)/OT 1.10 (0.05, 22.5)¶ 0.09¶ NI∥ – –
Any public toilet use (adult) 1.47 (0.44, 4.96) 0.53 0.33 0.42 0.43
High public toilet use (> 10 times per month, adult) 2.38 (0.36, 17.0) 0.36 0.75 0.64 0.24
Sanitation (cluster level)
High HH toilet coverage 3.32 (0.62, 21.2) 0.17 1.47 0.57 0.01
Low HH toilet coverage 1.05 (0.30, 3.68) 0.94 −0.51 0.43 0.24
High coverage of poor FSM 2.29 (0.31, 17.8) 0.41 2.50 1.04 0.02
High coverage of poor FSM/OT NI∥ – −2.34 1.26 0.07
CI = confidence interval; FSM = fecal sludge management; GII = genotype II; HH = household; NI = no interaction; OR = odds ratio; OT = Old Town; SE = standard error.
*Models are adjusted for neighborhood and hygiene status (“good” or “poor”, as discussed previously). Interactions of sanitation variable and neighborhood were tested for all models and
are presented if P < 0.10.
†Concentration differences are in log10 genome equivalent copies/100 mL.
‡Estimated relative to all other households, including those with toilets with other associated FSM and those without toilets.
§Estimated relative to households without a toilet or those with “other” FSM practices.
¶OR for interaction is presented as predicted OR for the OT neighborhood (from the model), not as the exponentiation of the interaction term alone. P value presented is for the interaction
term alone.
∥NI with neighborhood was included in this model (P ≥ 0.10 for interaction term).
 URBAN SANITATION AND FECAL CONTAMINATION 1411

to households with toilets compared with those without toi- DISCUSSION

lets, though these differences were not significant (P = 0.07
and P = 0.10, respectively; Table 4: Sanitation (household This study examined the associations between household
level)). At the cluster level, GII norovirus concentrations in sanitation—including toilets, FSM, and spatial heterogeneity
of sanitation coverage—and fecal contamination within the
drains within the cluster of high toilet coverage were signifi-
household and the local urban environment and enteric infec-
cantly higher (by 1.47 log10 GEC/100 mL, P = 0.01; Table 4:
tion in young children. The results suggest that the FSM,
Sanitation (cluster level)) than in the rest of the neighbor-
neighborhood-level coverage, and spatial clustering of
hood. Associations between clusters of poor FSM and
household toilets are significantly associated with household-
GII norovirus concentrations in drains varied by neighbor-
and neighborhood-level fecal contamination and pediatric
hood. In Chinnallapuram, GII norovirus concentrations were
enteric infection prevalence. Enteric infection was signifi-
significantly higher within the cluster of high coverage of
cantly more prevalent among children in households with
poor FSM than the rest of the neighborhood (difference in
poor FSM, even when compared with children in households
concentrations: 2.50 log10 GEC/100 mL, P = 0.02), whereas
without toilets, and was least prevalent among children in
in Old Town, there was no significant association (differ-
households with good FSM. In areas of high coverage of
ence in concentrations: 0.16 log10 GEC/100 mL, P = 0.07 household toilets, drains in clusters of poor FSM had higher
for interaction). concentrations of GII norovirus compared with the rest of
Household- and cluster-level sanitation and enteric the study area. In areas of low coverage of household toilets,
infection in children. Associations between household- the presence of a toilet was associated with higher E. coli
and cluster-level sanitation and concurrent enteric infection concentrations on household floors.
in children (detection of any enteric pathogen in children’s This study is one of the first to examine urban, household
stool) were evaluated by mixed-effects logistic regression toilets by both their spatial heterogeneity and associated FSM
(Table 5). Children in households with a toilet with good and describe associations with pediatric enteric infection.
FSM had 55% lower prevalence of infection, compared with Comparison of onsite excreta containment in tanks under the
children in households with toilets with poor FSM or no household to open drainage is new to the literature, which
toilet present (P = 0.17; Table 5: Household level). Con- has previously focused on sewerage.19,46,47 Although evi-
versely, children in households with a toilet with poor FSM dence of lower incidence of pediatric diarrhea associated with
had 3.78 times higher prevalence of infection when pooled urban drainage interventions exists, our findings suggest
across both neighborhoods (P = 0.05, data not shown), though that toilets that discharge to open drains may be potential
this association varied by neighborhood. In Chinnallapuram, risk factors for fecal exposure and enteric infection.48 Finally,
children in households with a toilet with poor FSM had this is one of the few studies to quantify fecal contamination
10 times the prevalence of enteric infection of children in (including fecal indicator bacteria and enteric pathogens)
other households (P = 0.04), whereas in Old Town, no signifi- in both the household and public domain as an outcome
cant association was present. Similar associations were and examine its associations with sanitation in low-income
observed when separating children by households with a urban areas.
toilet with good FSM, those with a toilet with poor FSM, or Provision of household toilets without regard for associ-
those without a toilet. Prevalence of enteric infection did ated FSM practices may still contribute to fecal contamina-
not vary significantly with cluster-level sanitation variables tion in the local environment and may not reduce pediatric
(Table 5: Cluster level). enteric infection when compared with the absence of a toilet,

TABLE 5
Any enteric pathogen detection in child stool* by household- and cluster-level attributes†
PR (95% CI) P value

Household level
Household toilet 1.57 (0.45, 5.49) 0.48
Toilet excreta contained onsite‡ 0.45 (0.14, 1.43) 0.17
Toilet discharges to drain‡ 10.0 (1.52, 200)¶ 0.04¶
Toilet discharges to drain/OT 0.58 (0.04, 7.94)¶ 0.03¶
Toilet excreta contained onsite§ 0.69 (0.12, 3.73) 0.67
Toilet discharges to drain§ 8.33 (1.02, 181) 0.08
Toilet excreta contained onsite/OT 0.66 (0.05, 8.62)¶ 0.98¶
Toilet discharges to drain/OT 0.53 (0.03, 8.61)¶ 0.05¶
Open defecation (< 5-year-old) 0.38 (0.10, 1.50) 0.17
Open defecation (adult) 0.83 (0.21, 3.32) 0.79
Any public toilet use (adult) 1.50 (0.54, 4.20) 0.44
High public toilet use (> 10 times per month, adult) 0.78 (0.16, 3.74) 0.76
Cluster level§
High cluster of household toilets 0.75 (0.17, 3.33) 0.71
Low cluster of household toilets 0.73 (0.26, 2.09) 0.56
High cluster of household toilets discharging to drain 2.55 (0.43, 15.1) 0.30
CI = confidence interval; OT = Old Town; PR = prevalence ratio.
*Pathogens detected in stool included astrovirus, Campylobacter spp., Entamoeba histolytica, Giardia spp., genotype II norovirus, and pathogenic Escherichia coli.
†Models with PRs for detection of any enteric pathogen in stool specimens presented, adjusted for neighborhood and hygiene status (“good” or “poor”).
‡Estimated relative to all other households, including those with toilets with other associated fecal sludge management (FSM) and those without a toilet.
§Estimated relative to households without a toilet or those with “other” FSM practices.
¶Odds ratio (OR) for interaction is presented as predicted OR for the OT neighborhood (from the model), not as the exponentiation of the interaction term alone. P value presented is for
the interaction term alone.
1412 BERENDES AND OTHERS

especially in dense, urban areas. High coverage of toilets suggested that shared sanitation is associated with lower
discharging directly into the local environment combined with fecal contamination in the toilet itself.58
ubiquitous presence of open drains was associated with Associations between household toilet presence and
higher concentrations of human-specific pathogens, for increased fecal contamination in low-coverage areas under-
example, GII norovirus, in drains in this study.49 Although score the importance of sanitation at the community level, in
structured observations from other low-income, urban areas addition to the household level. Evidence suggests sanitation
suggest contact with drains is not common for this age coverage—including household toilets and FSM—must sur-
group, QMRAs suggest that enteric infection may occur after pass a threshold before community-level fecal contamination
a single contact.14,15,50,51 Further, children likely have fre- is sufficiently reduced to confer health benefits.16,22,59 Nota-
quent contact with the floors and ground both inside and out- bly, in previous studies of high coverage of rural sanitation,
side the household, as well as contact with caregivers or these health benefits were not limited to only those house-
other family members, all of which may provide direct or indi- holds with toilets.23,24,60
rect exposure to feces from the outside environment.50,52 Although this study included more detailed sanitation
Thus, though the specific exposure pathway is unclear, there exposures, including multiple FSM typologies and spatial
are multiple pathways by which uncontained excreta regularly heterogeneity, it is limited in its assessment of some potential
discharged from nearby toilets may pose a risk to young chil- confounders, including hygiene practices and exposure to
dren in urban environments.16 animals. Classification of households into “good” or “poor”
Onsite containment of feces, if properly desludged and hygiene based on a 50% score cutoff may have misclassified
transported away after filling, is expected to yield lower levels those with scores around the cutoff. Further, the household
of environmental contamination, and subsequently lower hygiene score included questions about practices at the
prevalence of enteric infection. Children in these households household toilet, limiting our ability to separate conclusions
had the lowest prevalence of enteric infection in this study about sanitation and hygiene practices since households
when compared with those in households without toilets or without a toilet were more likely to have lower scores. Fur-
with toilets but with poor FSM. This finding suggests that ther, questions about interaction with animals were not part
fecal contamination—and therefore exposures in the local of the household hygiene score or the household survey,
environment—were lower, though significant differences in thus the contributions of animals to the spread of fecal con-
drain concentrations were not observed. Though there is tamination were not measured.
systematic evidence showing that sewerage interventions, While a census was not feasible, household selection
another form of contained FSM, are associated with decreased approximated the spatial distribution of households by random
incidence of diarrhea and enteric infection, onsite contain- selection within each of five sub-neighborhood areas. Using
ment with emptying and transport has not been studied.19 this approach and estimating clusters by spatial scan provided
Further, in contrast to our findings, recent city-level assess- a more accurate assessment of the spatial heterogeneity
ments of FSM have shown that long-term management underlying the neighborhood, which is significant with regard
of onsite fecal sludge is generally poor and may lead to to socioeconomic variables in low-income, urban settings.61
nonfunctional toilets and backflow of excreta into homes The number of environmental samples and stool speci-
during floods, contaminating the household environment.7,17 mens collected limited the study’s ability to conduct analy-
Given that we did not have the ability to assess the frequency ses on further subgroups of sanitation within neighborhoods.
or quality of FSM longitudinally, we cannot definitively make However, assessment of fecal indicator bacteria and patho-
conclusions about the long-term effectiveness of the onsite gens in environmental samples and enteric infection in stool
containment in place. This is an important knowledge gap for provided more comprehensive and objective outcomes than
future sanitation and health studies, given the increasing fre- previous measures of self-reported diarrhea.12,30 While
quency with which onsite excreta containment with trucking detection of pathogens was possible in all cases, inconsis-
is observed in urban areas and the prohibitive costs, plan- tent standard curves for EAEC in drain samples prevented
ning, and operation and maintenance logistics associated estimation of the CE quantities present, limiting power to
with sewer systems.53 detect differences. Although the low prevalence of patho-
In areas of low coverage of household toilets, their pres- gens in samples within the household limited analyses of
ence was associated with increased fecal contamination on E. coli, this difference in detection between organisms is
household floors, which may reflect sharing of toilets and important when considering use of fecal indicator bacteria
subsequently poorer maintenance and hygiene. Although against use of fecal pathogens.
our study did not collect information on households’ shar- Consideration of sanitation beyond the household toilet,
ing of toilets, the practice is not uncommon among house- including the household’s FSM within the neighborhood envi-
holds without sanitation in poor urban areas of India.54 A ronment, is necessary to better understand how reductions
recent study of peri-urban homes in Peru indicated house- in pediatric enteric infection may be achieved in urban areas.
hold shared sanitation was associated with increased fecal Despite new efforts to diagnose FSM conditions, the effects
contamination on kitchen floors when compared with indi- of FSM typologies on fecal contamination and enteric infec-
vidual household sanitation, and household shared sanita- tion are not well described to date, and open drains persist
tion has previously been linked to increased diarrheal as default sewerage options throughout many low-income
prevalence in children.55–57 However, though it is assumed countries.7,17,62 At a minimum, future studies should quantify
that shared sanitation yields higher household-level fecal the effects of this hazard at the household, neighborhood,
contamination through poor toilet maintenance, further and city scales.7
research is necessary as general toilet maintenance was Overall, this study provides evidence of the importance
not measured in our study, and recent evidence has of both FSM and the spatial heterogeneity of household
 URBAN SANITATION AND FECAL CONTAMINATION 1413

toilets within neighborhoods when evaluating the effective- 4. Alirol E, Getaz L, Stoll B, Chappuis F, Loutan L, 2011. Urbani-
ness of urban sanitation. To reduce fecal contamination sation and infectious diseases in a globalised world. Lancet
Infect Dis 11: 131–141.
and improve health, good FSM must accompany increases 5. Mara D, Lane J, Scott B, Trouba D, 2010. Sanitation and health.
in toilet coverage to remove and safely contain sewage PLoS Med 7: e1000363.
from the urban public and private domains. Because iso- 6. Baum R, Luh J, Bartram J, 2013. Sanitation: a global estimate
lated toilets in low-coverage areas may not yield the same of sewerage connections without treatment and the resulting
benefits as clusters of toilets in high-coverage areas, sani- impact on MDG progress. Environ Sci Technol 47: 1994–2000.
7. Peal A, Evans B, Blackett I, Hawkins P, Heymans C, 2014.
tation must be considered at the community level. As urban Fecal sludge management: a comparative analysis of 12 cities.
populations continue to grow, sanitation interventions must J Water Sanit Hyg Dev 4: 563–575.
contain excreta in both the public and private domain to 8. Isunju JB, Schwartz K, Schouten MA, Johnson WP, van Dijk
protect health and the environment. MP, 2011. Socio-economic aspects of improved sanitation
in slums: a review. Public Health 125: 368–376.
9. Victora CG, Smith PG, Vaughan JP, Nobre LC, Lombardi C,
Received March 3, 2016. Accepted for publication December 2, 2016. Teixeira AM, Fuchs SC, Moreira LB, Gigante LP, Barros FC,
Published online March 20, 2017. 1988. Water supply, sanitation and housing in relation to the
risk of infant mortality from diarrhoea. Int J Epidemiol 17:
Note: Supplemental information and tables appear at www.ajtmh.org. 651–654. http://www.ncbi.nlm.nih.gov/pubmed/3209344.
Acknowledgments: We would like to thank the field and lab staff at 10. D’Souza RMD, 1997. Housing and environmental factors and
CMC in Vellore, including Sheela Roy and field teams, for their effort their effects on the health of children in the slums of Karachi,
on this study and all aspects of the SaniPath study in Vellore, India. Pakistan. J Biosoc Sci 29: 271–281.
Further, we would like to thank the Bill & Melinda Gates Foundation for 11. Katukiza AY, Ronteltap M, Niwagaba CB, Foppen JWA,
Kansiime F, Lens PNL, 2012. Sustainable sanitation technology
funding this study and the two anonymous reviewers for their careful
options for urban slums. Biotechnol Adv 30: 964–978.
commentary and suggestions that greatly improved this manuscript.
12. Clasen T, Bostoen K, Schmidt W, Boisson S, Fung I, Jenkins
Financial support: This study was funded by the Bill and Melinda M, Scott B, Sugden S, Cairncross S, 2010. Interventions to
Gates Foundation. improve disposal of human excreta for preventing diarrhoea
(review). Cochrane Database Syst Rev CD007180.
Disclaimer: The funders had no role in the design or implementation 13. Wolf J, Prüss-Ustün A, Cumming O, Bartram J, Bonjour S,
of this study Cairncross S, Clasen T, Colford JM, Curtis V, De France J,
Authors’ addresses: David Berendes, Department of Environmental Fewtrell L, Freeman MC, Gordon B, Hunter PR, Jeandron A,
Engineering, School of Civil and Environmental Engineering, Georgia Johnston RB, Mäusezahl D, Mathers C, Neira M, Higgins
Institute of Technology, Atlanta, GA, and Center for Global Safe JPT, 2014. Assessing the impact of drinking water and sanita-
Water, Sanitation, and Hygiene, Emory University School of Public tion on diarrhoeal disease in low- and middle-income settings:
Health, Atlanta, GA, E-mail: [email protected]. Amy systematic review and meta-regression. Trop Med Int Health
Kirby, Suraja Raj, Habib Yakubu, Juan Leon, and Katharine Robb, 19: 928–942.
Center for Global Safe Water, Sanitation, and Hygiene, Emory Uni- 14. Labite H, Lunani I, van der Steen P, Vairavamoorthy K, Drechsel
versity School of Public Health, Atlanta, GA, E-mails: aekirby@emory. P, Lens P, 2010. Quantitative microbial risk analysis to evaluate
edu, [email protected], [email protected], juan.leon@ health effects of interventions in the urban water system of
emory.edu, and [email protected]. Julie A. Clennon, Department of Accra, Ghana. J Water Health 8: 417–430.
Biostatistics and Bioinformatics, Emory University School of Public 15. Katukiza AY, Ronteltap M, van der Steen P, Foppen JWA, Lens
Health, Atlanta, GA, and Center for Global Safe Water, Sanitation, PNL, 2014. Quantification of microbial risks to human health
and Hygiene, Emory University School of Public Health, Atlanta, GA, caused by waterborne viruses and bacteria in an urban slum.
E-mail: [email protected]. Arun Kartikeyan, Priya Hemavathy, J Appl Microbiol 116: 447–463.
Annai Gunasekaran, and Ben Ghale, Wellcome Research Laboratory, 16. Wagner EG, Lanoix JN, 1958. Excreta Disposal for Rural Areas
Christian Medical College and Hospital Vellore, Vellore, India, E-mails: dr. and Small Communities. Geneva, Switzerland: WHO.
[email protected], [email protected], annaimbavellore@ 17. Peal A, Evans B, Blackett I, Hawkins P, Heymans C, 2014. Fecal
gmail.com, and [email protected]. J. Senthil Kumar and Venkata sludge management (FSM): analytical tools for assessing
Raghava Mohan, Department of Community Health, Christian Medical FSM in cities. J Water Sanit Hyg Dev 4: 371.
College and Hospital Vellore, Vellore, India, E-mails: senthilcmc2008@ 18. Strande L, Ronteltap M, Brdjanovic D, (eds.), 2014. Faecal
gmail.com and [email protected]. Gagandeep Kang, Division of Sludge Management: Systems Approach for Implementation
Gastrointestinal Sciences, Wellcome Research Laboratory, Christian and Operation. London, United Kingdom: IWA Publishing.
Medical College Vellore, E-mail: [email protected]. Christine 19. Norman G, Pedley S, Takkouche B, 2010. Effects of sewerage
Moe, Emory Global Health Institute, Emory University, Atlanta, GA, on diarrhoea and enteric infections: a systematic review and
E-mail: [email protected]. meta-analysis. Lancet Infect Dis 10: 536–544.
20. Shuval HI, Tilden RL, Perry BH, Grosse RN, 1981. Effect of
This is an open-access article distributed under the terms of the investments in water supply and sanitation. Bull World Health
Creative Commons Attribution License, which permits unrestricted Organ 59: 243–248.
use, distribution, and reproduction in any medium, provided the 21. Bateman OM, Smith S, 1991. A Comparison of the Health
original author and source are credited. Effects of Water Supply and Sanitation in Urban and Rural
Guatemala. Washington, DC: United States Agency for Inter-
national Development.
REFERENCES 22. Root GPM, 2001. Sanitation, community environments, and child-
hood diarrhea in rural Zimbabwe. J Health Popul Nutr 19: 73–82.
1. Prüss A, Kay D, Fewtrell L, Bartram J, 2002. Estimating the 23. Eisenberg JNS, Trostle J, Sorensen RJD, Shields KF. Toward a
burden of disease from water, sanitation and hygiene at a systems approach to enteric pathogen transmission: from
global level. Environ Health Perspect 110: 537–542. individual independence to community interdependence.
2. Mbuya MNN, Humphrey JH, 2016. Preventing environmental Annu Rev Public Heal 33: 239–257.
enteric dysfunction through improved water, sanitation and 24. Emory University, 2014. SaniPath. Available at: www.sanipath.
hygiene: an opportunity for stunting reduction in developing org. Accessed November 21, 2015.
countries. Matern Child Nutr 12 (Suppl 1): 106–120. 25. John SM, Thomas RJ, Kaki S, Sharma SL, Ramanujam K,
3. UNICEF and World Health Organization, 2012. Progress on Raghava MV, Koshy B, Bose A, Rose A, Rose W,
Drinking Water and Sanitation: 2012 Update. New York: Ramachandran A, Joseph AJ, Babji S, Kang G, 2014. Estab-
WHO/UNICEF Joint Monitoring Programme for Water Supply lishment of the MAL-ED birth cohort study site in Vellore,
and Sanitation. southern India. Clin Infect Dis 59 (Suppl 4): S295–S299.
1414 BERENDES AND OTHERS

26. John J, Giri S, Karthikeyan AS, Iturriza-Gomara M, Muliyil J, able at: https://cran.r-project.org/package=logistf. Accessed
Abraham A, Grassly NC, Kang G, 2014. Effect of a single September 1, 2015.
inactivated poliovirus vaccine dose on intestinal immunity 43. Bates D, Maechler M, Bolker B, Walker S, 2014. Fitting linear
against poliovirus in children previously given oral vaccine: an mixed-effects models using lme4. J Stat Softw 67: 1–48.
open-label, randomised controlled trial. Lancet 384: 1505–1512. http://arxiv.org/abs/1406.5823.
27. Gladstone BP, Muliyil JP, Jaffer S, Wheeler JG, Le Fevre A, 44. Firth D, 1993. Bias reduction of maximum likelihood estimates.
Iturriza-Gomara M, Gray JJ, Bose A, Estes MK, Brown DW, Biometrika 80: 27–38.
Kang G, 2008. Infant mortality in an urban slum birth cohort. 45. Kulldorff M, 1997. A spatial scan statistic. Commun Stat Meth
Arch Dis Child 93: 479–484. 26: 1481–1496.
28. Sarkar R, Kattula D, Francis MR, Ajjampur SSR, Prabakaran 46. Barreto ML, Genser B, Strina A, Teixeira MG, Assis AMO, Rego
AD, Jayavelu N, Muliyil J, Balraj V, Naumova EN, Ward HD, RF, Teles CA, Prado MS, Matos SMA, Santos DN, dos
Kang G, 2014. Risk factors for cryptosporidiosis among Santos LA, Cairncross S, 2007. Effect of city-wide sanitation
children in a semi urban slum in southern India: a nested programme on reduction in rate of childhood diarrhoea in
case-control study. Am J Trop Med Hyg 91: 1128–1137. northeast Brazil: assessment by two cohort studies. Lancet
29. Houpt E, Gratz J, Kosek M, Zaidi AKM, Qureshi S, Kang G, Babji 370: 1622–1628.
S, Mason C, Bodhidatta L, Samie A, Bessong P, Barrett L, Lima 47. Kolahi A-A, Rastegarpour A, Sohrabi M-R, 2009. The impact of
A, Havt A, Haque R, Mondal D, Taniuchi M, Stroup S, McGrath an urban sewerage system on childhood diarrhoea in Tehran,
M, Lang D, 2014. Microbiologic methods utilized in the MAL-ED Iran: a concurrent control field trial. Trans R Soc Trop Med
cohort study. Clin Infect Dis 59 (Suppl 4): S225–S232. Hyg 103: 500–505.
30. Platts-Mills JA, Liu J, Gratz J, Mduma E, Amour C, Swai N, 48. Moraes LR, Cancio JA, Cairncross S, Huttly S, 2003. Impact of
Taniuchi M, Begum S, Yori PP, Tilley DH, Lee G, Shen Z, drainage and sewerage on diarrhoea in poor urban areas in
Whary MT, Fox JG, McGrath M, Kosek M, Haque R, Houpt Salvador, Brazil. Trans R Soc Trop Med Hyg 97: 153–158.
ER, 2014. Detection of Campylobacter in stool and determina- http://www.ncbi.nlm.nih.gov/pubmed/14584367.
tion of significance by culture, enzyme immunoassay, and 49. Okoh AI, Sibanda T, Gusha SS, 2010. Inadequately treated
PCR in developing countries. J Clin Microbiol 52: 1074–1080. wastewater as a source of human enteric viruses in the envi-
31. Liu J, Gratz J, Amour C, Kibiki G, Becker S, Janaki L, Verweij JJ, ronment. Int J Environ Res Public Health 7: 2620–2637.
Taniuchi M, Sobuz SU, Haque R, Haverstick DM, Houpt ER, 50. Teunis PFM, Reese HE, Null C, Yakubu H, Moe CL, 2016.
2013. A laboratory-developed TaqMan Array Card for simulta- Quantifying contact with the environment: behaviors of young
neous detection of 19 enteropathogens. J Clin Microbiol children in Accra, Ghana. Am J Trop Med Hyg 94: 920–931.
51: 472–480. 51. Gretsch SR, Ampofo JA, Baker KK, Clennon J, Null CA, Peprah
32. Kageyama T, Kojima S, Shinohara M, Uchida K, Fukushi S, D, Reese H, Robb K, Teunis P, Wellington N, Yakubu H, Moe
Hoshino FB, Takeda N, Katayama K, 2003. Broadly reactive CL, 2016. Quantification of exposure to fecal contamination in
and highly sensitive assay for Norwalk-like viruses based on open drains in four neighborhoods in Accra, Ghana. J Water
real-time quantitative reverse transcription-PCR. J Clin Microbiol Health 14: 255–266.
41: 1548–1557. 52. Perry S, De La Luz Sanchez M, Hurst PK, Parsonnet J, 2005.
33. Liu P, Hsaio H-M, Jaykus L-A, Moe C, 2010. Quantification of Household transmission of gastroenteritis. Emerg Infect Dis
Norwalk virus inocula: comparison of endpoint titration and 11: 1093–1096.
real-time reverse transcription-PCR methods. J Med Virol 53. Koné D, 2010. Making urban excreta and wastewater manage-
2010: 1612–1616. ment contribute to cities’ economic development: a paradigm
34. Collinet-Adler S, Babji S, Francis M, Kattula D, Premkumar PS, shift. Water Policy 12: 602–610.
Sarkar R, Mohan VR, Ward H, Kang G, Balraj V, Naumova 54. Heijnen M, Routray P, Torondel B, Clasen T, 2015. Shared
EN, 2015. Environmental factors associated with high fly den- sanitation versus individual household latrines in urban
sities and diarrhea in Vellore, India. Appl Environ Microbiol slums: a cross-sectional study in Orissa, India. Am J Trop
81: 6053–6058. Med Hyg 93: 263–268.
35. Brick T, Primrose B, Chandrasekhar R, Roy S, Muliyil J, Kang G, 55. Exum NG, Olórtegui MP, Yori PP, Davis MF, Heaney CD, Kosek
2004. Water contamination in urban south India: household M, Schwab KJ, 2016. Floors and toilets: association of floors
storage practices and their implications for water safety and and sanitation practices with fecal contamination in Peruvian
enteric infections. Int J Hyg Environ Health 207: 473–480. Amazon peri-urban households. Environ Sci Technol 50:
36. USEPA, 2002. Method 1604: Total Coliforms and Escherichia 7373–7381.
coli in Water by Membrane Filtration Using a Simultaneous 56. Fuller JA, Clasen T, Heijnen M, Eisenberg JNS, 2014. Shared
Detection Medium (MI Medium). sanitation and the prevalence of diarrhea in young children:
37. Gruber JS, Ercumen A, Colford JM, 2014. Coliform bacteria evidence from 51 countries, 2001–2011. Am J Trop Med
as indicators of diarrheal risk in household drinking water: Hyg 91: 173–180.
systematic review and meta-analysis. PLoS One 9: e107429. 57. Heijnen M, Cumming O, Peletz R, Chan GK-S, Brown J, Baker
38. Platts-Mills JA, Babji S, Bodhidatta L, Gratz J, Haque R, Havt A, K, Clasen T, 2014. Shared sanitation versus individual house-
McCormick BJ, McGrath M, Olortegui MP, Samie A, Shakoor hold latrines: a systematic review of health outcomes. PLoS
S, Mondal D, Lima IF, Hariraju D, Rayamajhi BB, Qureshi S, One 9: e93300.
Kabir F, Yori PP, Mufamadi B, Amour C, Carreon JD, Richard S a, 58. Exley JLR, Liseka B, Cumming O, Ensink JHJ, 2015. The
Lang D, Bessong P, Mduma E, Ahmed T, Lima AA, Mason CJ, sanitation ladder, what constitutes an improved form of
Zaidi AK, Bhutta Z a, Kosek M, Guerrant RL, Gottlieb M, Miller sanitation? Environ Sci Technol 49: 1086–1094.
M, Kang G, Houpt ER, 2015. Pathogen-specific burdens of 59. Briscoe J, 1984. Intervention studies and the definition of dom-
community diarrhoea in developing countries: a multisite birth inant transmission routes. Am J Epidemiol 120: 449–455.
cohort study (MAL-ED). Lancet Glob Health 3: 564–575. 60. Fuller JA, Villamor E, Cevallos W, Trostle J, Eisenberg JNS,
39. Okhuysen PC, Dupont HL, 2010. Enteroaggregative Escherichia 2016. I get height with a little help from my friends: herd
coli (EAEC): a cause of acute and persistent diarrhea of protection from sanitation on child growth in rural Ecuador.
worldwide importance. J Infect Dis 202: 503–505. Int J Epidemiol 45: 460–469.
40. Liu P, Hill VR, Hahn D, Johnson TB, Pan Y, Jothikumar N, Moe 61. Hacker KP, Seto KC, Costa F, Corburn J, Reis MG, Ko AIDiuk-
CL, 2012. Hollow-fiber ultrafiltration for simultaneous recovery Wasser MA, 2013. Urban slum structure: integrating socio-
of viruses, bacteria and parasites from reclaimed water. economic and land cover data to model slum evolution in
J Microbiol Methods 88: 155–161. Salvador, Brazil. Int J Health Geogr 12: 45.
41. R Core Team, 2015. R: A Language and Environment for Statis- 62. Jenkins MW, Cumming O, Scott B, Cairncross S, 2014. Beyond
tical Computing. Available at: https://www.r-project.org/. “improved” towards “safe and sustainable” urban sanitation:
Accessed June 18, 2015. assessing the design, management and functionality of sani-
42. Heinze G, Ploner M, Dunkler D, Southworth H, 2013. logistf: tation in poor communities of Dar es Salaam, Tanzania.
Firth’s Bias Reduced Logistic Regression: Version 1.2.1. Avail- J Water Sanit Hyg Dev 4: 131.